摘要
目的对比以R语言为基础结合BP神经网络和遗传算法与响应面分析2种方法,优化补阳还五汤(BHD)中主要有效成分的提取工艺。方法在单因素实验的基础上采用响应面实验设计方法,利用HPLC法检测提取得到的BHD中主要有效成分,计算其提取率。提取率结果进行熵权法赋值,得到综合评价值。在此基础上首先采用响应面分析方法得到其最佳提取工艺和综合评价预测值。再通过另一种优化分析方法:R语言环境下的BP神经网络和遗传算法,分别进行网络模型优化和目标寻优,以期得到另一组BHD中有效成分的最佳提取工艺和综合评价预测值。结果响应面处理方法得到的最优提取工艺为提取时间1.8 h、乙醇体积分数51%、提取温度91℃、液料比14∶1,该方法下综合评价预测值为908.45,验证试验平均值为897.58,相对误差为1.20%;R语言环境下BP神经网络和遗传算法处理得到的最佳提取工艺为提取时间2 h、乙醇体积分数40%、提取温度100℃、液料比14∶1,该模型综合评价预测值为907.71,验证试验平均值为905.33,相对误差为0.26%。结论对比2种分析方法发现,神经网络结合遗传算法的相对误差较小,与验证试验拟合度较高,即R语言环境下的BP神经网络和遗传算法数学模型可用来对BHD中有效成分的提取工艺进行分析和预测,优于响应面分析,为实现中药有效成分目标寻优以及中药现代化提供了新的思路和参考。
Objective To optimize the extraction process of main active ingredients from Buyang Huanwu Decoction(BHD) by comparing BP neural network combined with genetic algorithm under R language environment with response surface analysis. Methods On the basis of single factor test, the response surface design method was adopted and the main active ingredients of the extract were determined by HPLC. The results were presented in the form of extraction rate. The comprehensive evaluation value of the results was obtained by using the entropy method. Based on this, the best extraction process and the predictive comprehensive evaluation value of extraction rate were firstly obtained by using response surface analysis method. In order to find another best extraction process and comprehensive evaluation predictive value of the main active ingredients in BHD, the optimization of the network model and the discovery of the optimal target were conducted through the BP neural network combined with genetic algorithm under R language environment, respectively. Results The optimum extraction technology in response surface analysis were as follows: Extraction time was 1.8 h; Ethanol concentration was 51%; Extraction temperature was 91 ℃; Liquid material ratio was 14∶1. Under the condition, the comprehensive predicted value was 908.45; The average of the verification test was 897.58; The relative error was 1.20%. The best extraction process by BP neural network combined with genetic algorithm under R language environment was as follows: extraction time was 2 h; Ethanol concentration was 40%; Extraction temperature was 100 ℃; Liquid material ratio was 14∶1. Under the condition, the predicted value was 907.71; The average of the verification test was 905.33; And the relative error was 0.26%. Conclusion Through the comparison of the two analytical methods, it was found that the relative error of neural network combined with genetic algorithm is smaller and the fitting degree with the verification test is higher. That was to say, BP neural network model combined with genetic algorithm in R language environment was more suitable than response surface analysis to optimize the extraction process of the main active ingredients in BHD, which provided a new idea and reference for the discovery of active ingredients and modernization of Traditional Chinese Medicine.
出处
《中草药》
CAS
CSCD
北大核心
2018年第1期135-141,共7页
Chinese Traditional and Herbal Drugs
基金
国家自然科学基金重点项目(81630105)
国家自然科学基金面上项目(81374053)
浙江省自然科学基金重点项目(LZ17H270001)
关键词
补阳还五汤
提取工艺
响应面法
R语言
神经网络
遗传算法
HPLC
羟基红花黄色素A
芍药苷
阿魏酸
毛蕊异黄酮
芒柄花素
Buyang Huanwu Decoction
extraction process
response surface methodology
R language
neural network
genetic algorithm
HPLC
hydroxyl-saffior yellow A
paeoniflorin
ferulic acid
calycosin
formononetin